Models of search in labor markets are potentially of great use for policy analysis since their parameters are structural. However, a common feature of these models is that an assumption of optimal behavior on the part of agents is necessary to achieve identification. From a classical econometric perspective, this means the assumption of optimality is untestable and, if optimality is not imposed, it is impossible to learn about the unidentified parameters. This paper argues that Bayesian methods can overcome both of these problems. In particular, we discuss testing optimality in stationary job search models with reservation wages. Learning about economically meaningful quantities such as the discount rate and risk aversion, not identified by the data alone, is considered.
- reservation wage
- posterior simulation